Project 2: NBA All-Stars
Data Wranglers
Our Shiny App can be found here. Our source code can be found here (the shiny folder, specifically, holds our code, while Shiny.R is our complete App). Link if the hyperlink doesn’t work: https://github.com/vizdata-f21/project-2-data_wranglers.
A link to a screen recording for our presentation can be found here (YouTube).
Introduction
The NBA is an increasingly global league where its top talent comes from countries all over the world and colleges around the United States. As fans of basketball, we were very interested in looking at which teams All-Stars played for, where they were drafted, and how many NBA All-Stars (or the NBA’s best players) come from countries other than the United States and how international talent has grown over the years. This could also be used by NBA scouts and general managers to see how important international scouting is and whether they should focus a lot of their attention going into an NBA draft on foreign players.
Additionally, as Duke and college basketball fans, we wanted to explore which colleges had the most NBA All-Stars and which states had the most colleges with top NBA talent too. Our Shiny App enables users to analyze which colleges are most useful in producing NBA All-Star players, and where these colleges are found.
Data
This data comes from Basketball Reference, Real GM, Wikipedia, The United States Census Bureau, & The United States Department of Education. Data from Wikipedia, Basketball Reference, and RealGM was scraped by us. Prior to scraping data, we verified using the robots.txt file that the data was scrape-able.
Research Questions
Which colleges did the most NBA All-Star players attend? Geographically, which regions in the US are those colleges in that produce the most NBA All-Star players?
What proportion of NBA All-Stars were drafted in the first round of the NBA draft? Which teams did these All-Stars play for?
Where are NBA All-Stars originally from geographically? How many All-Stars, have there been from outside the United States per year? Which states have produced the most all stars per capita?
NBA All-Stars by College
Our first research question was primarily motivated by our group members being Duke basketball fans and fans of college basketball as a whole. We wanted to do a deeper dive into relationships between college basketball teams and NBA player success. In our analysis, we defined a player’s college as being the place that they attended immediately prior to starting their NBA career, regardless of whether they graduated.
In our Shiny App, we explored these relationships in the first three tabs: the “Colleges”, “College Map”, and “College State” tabs.
In the “Colleges” tab, we created a dynamic racing bar chart that measures the amount of All-Stars a college has produced since the introduction of the All-Star game in 1951. We allow the user to define whether they would like to count every appearance made in an All-Star game or to only count each player once per college, even if they made several All-Star games. We have other user options as well, as shown below.
To give us a broad sense of what colleges have been most successful in producing All-Star NBA players, we chose to plot the number of All-Star game appearances over time by college, yielding the result that Kentucky, UCLA, and UNC are the leaders of the pack.
One thing I should point out here is the use of University colors to give the plot more life (which took way more effort than I had anticipated).
To account for the fact that some players may have accounted for many All-Star appearances, we then limit the query to count just the number of players who have made an All-Star game, rather than the total amount of appearances. Unsurprisingly, this plot shows similar results, however Duke does leap UNC in cumulative All-Star players, undoubtedly due to the fact that Michael Jordan’s 14 appearances in the prestigious game are only counting as one tally in UNC’s column here.
One aspect of this data that we were keen to explore was the recent years. In 2006, the NBA instituted the “One-and-Done” rule, requiring players to be one year removed from high school before entering the NBA draft. This meant elite high school basketball players would be significantly more likely to spend at least one year in college prior to joining the NBA, contrasting the trend of players joining the NBA immediately following high school that had persisted in years prior. In the wake of this rule, colleges like Kentucky and Duke elected to pursue these elite athletes, valuing their basketball ability highly and dismissing concerns about the lack of continuity that would result from players coming and going yearly.
We chose to plot the same data as previously, limiting the starting year to be 2007, which was the first All-Star that “One-and-Dones” were in the NBA for.
In both total All-Star game appearances and number of players with an All-Star appearance, Kentucky has dominated the “One-and-Done” era, offering evidence for our hypothesis. Duke has had six All-Star players in the same era, good enough for second place. The effect of one dominant player can be seen clearly in the first chart, with Texas’ Kevin Durant’s 11 All-Star game appearances propelling Texas to a second place finish in the timespan. Kevin Durant was a “One-and-Done” player, drafted second overall in the 2007 NBA draft.
To find where NBA players attended colleges geographically, we plotted both a chloropleth map of states and a point location map showing where players have attended. Customizability for the user in the Shiny App is shown below.
We chose to use our Shiny App to examine the geography of where NBA All-Stars have gone to college over time.
We then chose to look on a bigger scale, comparing across states. To avoid problems that arise with red-green colorblindness, we chose a blue-red color palette, using the NBA logo’s colors for guidance.
California, North Carolina, and Kentucky stand out most among states. However, because states with many people in them would be more likely to have NBA All-Star players attend college there, we wanted to standardize states for comparison by population.
After normalizing states, Kentucky and North Carolina are still among the leaders. These states feature heralded basketball programs from which many great NBA players have emerged. Wyoming and Montana also stand out, but their values are more a product of their minuscule populations.
NBA All-Stars by Draft, Team
For our second question, we wanted to see how important early picks in the NBA draft are for finding the best players. We created an animated, racing bar chart to illustrate the number of All-Stars that were drafted in the first round, second round, or went undrafted and how that has shifted over the years. This could be used by NBA front offices to determine how important a first round pick is and how much a second round pick is really worth in trade scenarios. It was also interesting for us to look at how many undrafted All-Stars there have been, as we hypothesized that would be a very small percentage of the NBA’s top talent.
Another hypothesis we had was that a lot of the NBA’s top talent comes from the same few organizations over the years (such as the Boston Celtics and LA Lakers). We were interested in exploring the number of All-Stars per team over the years to draw conclusions on the distribution of talent in the NBA.
We also wanted to see which teams have had the most cumulative All-Star selections, so we created an interactive plot to address the second part of this research question. This was interesting for our group as we hypothesized that there would be a significant difference in the number of All-Stars from larger markets (like Los Angeles, New York, and Miami) versus small market cities (such as Memphis or Minneapolis). We wanted to analyze if the location of cities plays a significant in attracting “All-Star” talent to the same few teams.
First, for the “Draft Positions” tab to clean up the data I had to create a new variable called “draft_range” that used the str_detect function to determine whether each NBA All-Star was a first round pick, second round pick, or went undrafted. I also created a new variable called “sum_all_stars” which I used to count how many All-Stars there were from draft category each year.
Our Shiny App displays the distribution of which NBA draft picks cumulatively (over a range of years) are from each round of the draft or are undrafted. The distribution over the full range of years is illustrated below:
As you can see above, from 1951 to 2021 there have been significantly more total All-Star appearances from first round picks compared to both second round picks and undrafted players. There have been ~1300 cumulative first round pick All-Star selections compared to ~200 from the secound round and ~150 undrafted All-Stars. This distribution demonstrated the importance of earlier draft picks as this plot shows that over 75% of All-Star selections come from the first round of the draft. This outcome could encourage the idea of “tanking” as, if bad teams lose on purpose, they could get earlier draft picks. However, despite the large majority of All-Star selections being from the first round, second round picks are often basically traded away for little to no value and do not hold much trade value. Our group concluded that NBA general managers should still hold some value on those picks, as, we felt that the distribution is large enough to demonstrate that those picks are worth holding onto. We concluded that it is also worth it for teams to consider picking up undrafted players, they are almost 10% of the NBA’s All-Stars over the last 70 years and are often on much smaller contracts and are, therefore, low risk investments. However, more recently (as you can see by adjusting the range of years on the app) the proportion of second round picks and undrafted All-Star selections has decreased. For example, since 2010 there has not been any undrafted All-Star selections, which could signify that NBA teams are doing a better job drafting and undrafted players have less value than they did previously.
Next, for the “Team” tab, we looked at the distribution of total All-Star appearances for each team over a range of years. First, as many teams have changed cities and/or names over the course of the last 70 years, I used the mutate variable to rename each team by their current name and city. For example, the Los Angeles Lakers used to be the Minneapolis Lakers, so I used the mutate function so the total All-Stars would all be shown under the name “Los Angeles Lakers”. I also, similarly, created a new variable called “sum_all_stars” which I used after using the pivot functions to count how many All-Stars there were from draft category each year. Finally, for this chart I added a variable called “number_to_rank” that would allow the user to choose the number of teams he wants displayed. The default is set at all 30 teams, but if the user wanted a closer look at the teams with the most All-Stars, you could change the teams to range to 10, for example, and the app would only display the teams with the 10 most All-Stars cumulatively each year. The distribution over the full range of years is illustrated below:
All-Star appearances by NBA team
As you could see in the plot above, the larger market teams (from bigger, more attractive cities) have had pretty significantly more All-Star selections than most of the smaller market teams. Four of the top five teams with the most total All-Star appearances are from bigger markets in Boston, LA, Philadelphia, and New York. The teams with the lowest amount of total All-Star appearances are from smaller market cities, including Memphis, Charlotte, New Orleans, and Minnesota.
The total All-Star appearances could be explained by bigger markets, but also by managers building successful teams that resulted in more championships. The Lakers and Celtics have not only had the most All-Stars but they have had the most NBA championships. This is likely due to a combination of team and draft success, as well as their cities being attractive enough to bring in the best NBA players.
Question 3
Our third research question was motivated by a curiosity to not only understand the recent surge in international talent in the NBA, but also to identify which states were more likely to produce an NBA All-Star than others.
The tabs in our shiny app that correspond to this question are the “World Map” tab and the “U.S. Map” tab. For all the visualizations in these two tabs, we have filtered the data to only include one appearance per player, so players aren’t counted twice. We started by cleaning the data. The original All-Star NBA data had a nationality value for each player, but this value often had multiple values in it. This was because players may have been born abroad but grew up or studied in the United States. We cleaned up this column and then joined this data with U.S. states data for the same players to have one table that included players’ nationalities and their city and state of birth, if born in the United States.
For the world map, we plotted the chloropleth using ggplot2 default mapping data then joined this data with our All-Star data to get counts for the number of All-Stars from different countries. Here is an example visualization below:
The year range is set to the maximum, 1951-2021, but this is adjustable. The polygons representing countries on the map are filled in on a log scale, with blue representing the lower end of the scale and red the upper end. These colors match the colors in the NBA logo. The log scale was necessary as the United States has produced many more All-Stars than any other country, as expected. But, in the visualization below from 1951-1990, we can see that very few international countries had produced All-Stars.
This comparison between the two visualizations helps us see that the NBA has become a more international league recently, and better players from Europe, Africa, Asia, and other countries are starting to make a real impact on the league. Even still, there are very few international All-Stars, and we expect there to be more in the coming years, as scouts and teams use data like this to find international stars. One thing we found interesting is that there has never been a Canadian-born NBA All-Star, although this also may change in the coming years with Canadian talent like Andrew Wiggins and Shai Gilgeous-Alexander.
Now, moving to the U.S. Map tab, our motivation was to map the number of All- Stars that each state has produced, not by their college state but by their birth state. The tab has an option for whether the data should be shown “Per million residents,” and also a slider for the year range of the data.
The first visualization below is not per capita, just the pure number of All- Stars by birth state.
We can see from the visualization that many states in the Midwest have never produced All-Star talent, and this may be because basketball is not as popular at young ages in those states. But, states like California and New York, who have a historical basketball culture and immensely large populations, have produced more all stars than any other state. This may also be because of these states being home to “prep schools,” which are high performance athletic high schools for elite athletes. Many athletes born in these states choose to go to prep schools near home, and this prepares them better for college and the NBA.
The next visualization is All-Stars per capita, or Per million residents.
This visualization shows that California and New York are actually not producing the highest number of All-Stars per million residents. The states that have produced the most All-Stars per million residents are Louisiana, Arkansas, and Kentucky. These states have smaller populations but also a strong basketball culture, and this contributes to them producing high level NBA talent.
Overall Conclusions
In examining colleges, we found that Kentucky and Duke have emerged in recent years as All-Star producing powerhouses. After the introduction of the “One-and-Done” rule in 2006, both schools have elected to pursue elite high school basketball players which has enabled them to see their former players thrive in the NBA. UNC and UCLA are other schools that have produced many All-Star players, which is reasonable, as both schools have had immense success throughout their histories.
When examining draft positions, we found that, as expected, players drafted in the first round make up a much larger proportion of All-Stars than either those drafted in the second round or undrafted players. But, what we found interesting was that there have been almost the same number of undrafted All-Stars as those drafted in the second round. This is why, recently, we have seen many players choose to go undrafted rather than slip to the second round and wait for an opportunity that suits them. Looking at All-Stars by team, we visualized what we expected - big market teams that have been around for a long time have had many more all stars than smaller market teams, some of which were established more recently.
Finally, examining birth countries and states of All-Stars, we can see that, recently, there has been a surge in the number of international All-Stars as the league scouts more talent from abroad. We expect there to be even more international All-Stars in the coming years and beyond. When looking at birth states, we see that large states with historical basketball culture and many prep schools, like California and New York, have produced the largest number of All-Stars, but smaller states like Louisiana, Arkansas, and Kentucky have produced more All-Stars per capita.